Embedded Executive Podcast Embedded Executive: We Are In a Memory Crisis | Everspin
Apr 1, 2026
Sean Dougherty, Vice President at Everspin Technologies, is an executive focused on MRAM and non-volatile memory solutions. He discusses a real memory crisis driven by AI's massive capacity and bandwidth needs. He explains why this shortage is different, who gets prioritized for limited supply, and how MRAM alternatives and geopolitical forces shape short-term outlooks.
AI Snips
Chapters
Transcript
Episode notes
AI Demand Is Causing A Real Memory Shortage
- The current memory shortage is real and driven largely by skyrocketing AI demand for HBM, DRAM, and NAND.
- Memory makers are reallocating capacity to AI customers, creating tight supply and rising prices across the market.
Memory Has Become The Performance Bottleneck
- AI architectures need vastly more memory capacity and bandwidth, turning memory into a critical bottleneck for compute advances.
- NVIDIA, AMD, and Intel platforms increasingly rely on HBM and DRAM, and required networking to move data exacerbates the pressure.
Shortage Is Genuine Not Manufactured
- The shortage is genuine, not a coordination trick by suppliers; deployed compute is being fully utilized immediately.
- Unlike unused dark fiber in prior booms, newly deployed AI compute gets turned on and consumes memory right away.